The ability to capture and interpretelectrical impulses of brain neurons can radically affect the treatment of many diseases that do not currently have effective cure methods. A group of scientists from the University of California at San Francisco has developed a method for decoding the recording of periodic electrical activity of the cerebral cortex (electrocorticograms), which is carried out by obtaining data from electrodes placed directly on the surface of the brain.
Four people took part in the experiment.suffering from epilepsy, in the brain of which electrodes were implanted. The subjects were to read the prepared text aloud and make special offers. At the same time, instruments recorded the activity of the cerebral cortex.
The received data was transmitted to the neural network,where a detailed analysis was performed to identify spoken phrases by brain activity. Artificial intelligence associated, for example, the utilization of various sounds with certain movements of the patient's mouth.
The received data from the AI were transferred to the nexta neural network, which should have been deciphered by brain activity indicators, what the patient wanted to say. From 30 to 50 offers were checked.
The effectiveness of such speech recognition turned out to bevery high, and its accuracy exceeded the frequency of the natural reservations present in the conversation. The recognition accuracy of words by electrical impulses was 97%, while on average a person makes mistakes with reservations at least 5%.